CRT film-outs

[Paul Bruening]

Hey fellas,

 

I've got another goofy idea. I've got these eight workstations and am lousy with these 2K, true flat, 21" CRTs. My thinking is that if the quality of image is good enough, I could shoot them at around 2 fps with print stock and generate one first generation print per day per station. Do you think that a 1st gen 2K release print off of a high res CRT would look good stacked up against an industry standard 4th gen release print from something like a 4K Arrilaser neg?

 

That 4th gen release print carries only about 1.5K to 3/4K of the original's image information. So a 1st gen release print at 2K might hold up. I just don't know how those 2K CRTs image out in the first place. Has anyone seen footage shot off a 2K CRT?

 

I could rig them vertically, thus squeezing them into my boiler room. Mitchell NCs are going for $2,000 or less in these wretched economic times. I could make the 2000 ft mags from aluminum plate (or wood for that matter) using parts out of cheap 1000ft Mitchell mags.

 

If this worked out okay, I could conceivably knock out eight of these release prints per day (240 per month) using a contact printer to add the sound track to each.

 

Crazy or viable?

[Phil Rhodes]

You don't really want to use colour CRTs to do this sort of work, because it makes you subject to all the geometry and convergence problems they have. If you can find a mono CRT (often available for medical work) and do filtered exposures, you may still have geometry problems, but it's usually reasonably straightforward to get those down to an acceptably low level. What you won't get is convergence problems.

 

I also wonder about what "true flat" means in this context. Sometimes that means that the outside front of the tube is flat but the inside isn't, and I hate to think how you'd deal with trying to flat-field focus an image that's curved in exactly the wrong direction.

 

But yeah, I guess. I'd almost rather use a decent hi-def TFT, notwithstanding the lack of dynamic range, which are dead flat and have perfect geometry, althuogh you can probably achieve higher resolution with CRTs.

 

P

[Paul Bruening]

You don't really want to use colour CRTs to do this sort of work, because it makes you subject to all the geometry and convergence problems they have. If you can find a mono CRT (often available for medical work) and do filtered exposures, you may still have geometry problems, but it's usually reasonably straightforward to get those down to an acceptably low level. What you won't get is convergence problems.

 

I also wonder about what "true flat" means in this context. Sometimes that means that the outside front of the tube is flat but the inside isn't, and I hate to think how you'd deal with trying to flat-field focus an image that's curved in exactly the wrong direction.

 

But yeah, I guess. I'd almost rather use a decent hi-def TFT, notwithstanding the lack of dynamic range, which are dead flat and have perfect geometry, althuogh you can probably achieve higher resolution with CRTs.

 

P

 

Thanks for replying, Phil,

 

Can we mull on this one some more? These high res CRTs have a genuinely flat screen surface instead of the usual convex curve. They also have a full menu control system that can adjust every aspect of the display geometry. One of these View Sonic monitors is a video production monitor and has component inputs. Though, I don't have any component-out cards to route that way.

 

Can you help me understand your convergence concerns? Did I just address them?

 

I could do three separate RGB exposures of each frame and bypass the mechanics of a filter wheel. Does that sound doable?

 

My biggest concern is image quality. How will it look when the print ends up on my local Malco Theater's screen? If no one has seen a sample of this I'll just have to shoot a test.

[Phil Rhodes]

I uhoh.

 

You'll have to try it, although I'd be prepared to use some stupefyingly extreme gamma curves to get it to look anything even remotely like right, even if you pay close attention to matching the white point of the monitor to the film (CT meters and mired shift calculations here we come).

 

Convergence in a CRT monitor is a matter of getting the red, green and blue electron guns lined up. Because their position is not coincident, the manufacturing process is not completely consistent, because the earth and other nearby objects have a magnetic field, and because varying amounts of high-voltage drive are required to produce appropriate responses in each colour component, it's necessary to make often quite involved adjustments to precisely superimpose the three images and ensure that their geometry is rectilinear and consistent across the display area. This is invariably a process involving some compromise and it's rarely much fun - you're presumably aware of how magnetic phenomena are "squishy"; well, imagine having a dozen or more controls per channel and having to squish the three images around until they lines up to sufficient tolerance.

 

But even before you do any of that you have to address purity, which is a characteristic controlled by the need to fire each electron gun through some sort of mask in the front of the tube such that the red gun can only see red phosphor, etc. This is usually controlled by physically moving the magnetic coils around on the neck of the tube, or sometimes by adjusting small trimmer permanent magnets until a full-red field appears only and consistently bright red.

 

Neither of these are really user adjustments (especially moving the coils, which are generally cemented to the tube at manufacture). You may find there's convergence controls in the menu system of a recent CRT, which might help.

 

Once you get to this point you need to address moire concerns, which is the characteristic of CRT monitors that are being run at very high resolutions whereby the pixel pitch of the image is close in size to the mask inside the tube. The bigger the tube, ideally, the better you should do here.

 

The reason people use mono CRTs is that they're not really subejct to any of these issues other than having reasonably good geometry and focus, which is massively simpler than trying to achieve those things, in the same way, across three disparate pieces of hardware. Purity (other than flat-field evenness) and convergence are not an issue, geometry is massively simplified and moire does not occur.

 

I'm not sure what you're concerned about with regard to component inputs. Presumably these high-res CRTs are VGA monitors, which implies an RGB signal.

 

As to how good it'll look you'd probably have to try it even in the presence of someone else's experience. I suspect it'll be enormously dependent on exactly the hardware you have to hand. What I'd do would be to shoot a lot of single-frame test series using a wide variety of computer, monitor, camera, exposure and filtration setups and look at them on a light box. I suspect you'll end up using some really brutal LUTs to make it look good, so I'd make sure that these tests covered a very large range of settings, including things that look utterly wrong, because I think it'll probably end up looking utterly wrong through the viewfinder in order to look anything like right when projected. I suspect you may struggle to get much light output from the monitor to expose the intermediate stocks you may be considering, and you may need reasonably long exposures possibly with either much reduced red and green output on the monitor, or heavy blue filtration. CRTs are not generally very good at blue, especially if the blue has not been defocused to provide more output by exercising a larger area of phosphor.

 

P

[Paul Bruening]

Thanks, Phil,

 

That's exactly the kind of info I was hoping for. Separating the exposures into individual RGB might do something for the convergence problem. It may make LUTs and color tendencies of film and CRT easier too since forcing a single RGB image into extremes sounds like a stretch.

[John Sprung]

You might not have to make 2000 ft. mags -- I know we had them on the old PSR's for sitcoms, there may be some Mitchell ones out there, too. Look around before you expend too much effort on the build.

 

As for shooting off the CRT's you already have, first thing try some tests on a static image with a still camera. That'll tell you where your problems lie.

 

 

 

 

-- J.S.

[Phil Rhodes]

Separating the exposures into individual RGB might do something for the convergence problem.

 

I'm not sure it would. If you copy (for instance) the blue channel into the red and green channels on a colour CRT, then filter it so the film only sees the blue, you're back where you started. Only if you actually get a monochrome CRT do you get the convergence advantage.

 

That said I don't think convergence will be the biggest problem you hit. It can be tweaked to reasonable levels. I think the biggest issue you'll face will be getting enough output to allow you to shoot reels out in a reasonable amount of time, and getting the colour and contrast to look reasonable.

 

Another thought occurs. You will need to ensure you image a precise number of monitor scans per frame, which means either displaying the image for a certain number of complete scans using software (wherein the "black" level of the CRT may microscopically fog your image) or by syncing the shutter to the monitor and making sure it closes very quickly. What you can't do is just open the shutter and throw the image at the display, because most computer graphics software does not implement synchronisation between the display memory update and the monitor scan - if you update display memory half way down the scan, as is entirely feasible, you end up with one frame being output that has the bottom half full of picture and the top half empty. This will obviously cause flickering.

 

"Real" CRT recorders control the tube very precisely, and usually scan the image out several times (to increase output and dynamic range) in a visible set of triple flashes.

 

P

[Paul Bruening]

Interesting points. My thinking on the RGB has evolved into splashing digital color separations in sequence onto each frame not using any form of mechanical filtration at all. Basically, building the image by piling on three whole, unique color versions. I assumed I could arrive at a way to manage them to work with the print stock and balance out color-wise as well as all other image quality factors not related to the limits of a CRT's inherent grid pattern. I also assumed I could find a way to control their screen time precisely with software. That and a capping shutter along with a stepper motor would control all of the related timing matters and eliminate all of those dreaded kinescopic problems. In effect, run it as an animation stand letting the computer do all the boring control and job parts.

 

Frankly, I'd never use this system to produce more than a dozen prints for limited, marketing based release anyway. One rig taking a month or two to achieve that limited print quantity goal would be no problem. As long as the computer and rig are doing all of the work while I sit around watching TV and picking my nose, I'm fine.

 

Please, continue to help me with the flaws in my idea so I don't waste time chasing something that can't work no matter how I try to think around it. I really do appreciate the help, here, Phil.

[Phil Rhodes]

Interesting points. My thinking on the RGB has evolved into splashing digital color separations in sequence onto each frame not using any form of mechanical filtration at all.

 

I don't even think you need to do it sequentially. You won't gain yourself much. It gets complicated at this level because of another factor affecting CRT performance that I thought off at about 3am: power supply regulation. As we've seen, both geometry and brightness are affected by the load on the high voltage power supply, which increases nearly linearly with the amount of total light you're asking the tube to put out. It is technically difficult to make an EHT power supply in the tens-of-kilovolts range that is very load stable. For this reason you may wish to set the CRT to a fairly modest (even downright dim) level of output, which is probably exactly what you don't want to do, but which will moderate the load on the high voltage drive and minimise these issues.

 

Part of your tests should probably be a couple of hundred frames of the screen wiping from black to white - aim a spot meter at the white bit and see how much it dims as the screen fills with white. CRT-based video projectors, which are a bit of a special case as they demand enormous output from their tubes, are notorious for this issue - it isn't unusual for a 1/16-screen white block to be quite literally twice the brightness of a full white field, although most desktop CRTs won't suffer nearly so badly.

 

I assumed I could arrive at a way to manage them to work with the print stock and balance out color-wise

 

I suspect you will end up arriving at a look up table (perhaps even just a photoshop arbitrary map) which looks right, based on tests. If you are a C programmer, you'll then write code to apply that LUT to the image as it is displayed. If you're not, you'll batch process the reel through Photoshop before shooting it out. I've been on at the mplayer guys to add at least 1D LUTs to mplayer for a while.

 

I also assumed I could find a way to control their screen time precisely with software[/

 

I suspect you won't quite be able to do it with off the shelf software. Most video playback software (which is quite amenable to playing back in a single-frame-advance mode) does not provide any way to ensure that each video frame is played back for a precise number of monitor scans, even when the monitor refresh rate is an integer multiple of the desired frame rate. Graphics boards and their driver software provide a vertical retrace counter which allows you to count scans in software, and an interrupt when a scan is about to commence so you can flip buffers, but you would need to write C code to get access to this functionality.

 

Video playback software like mplayer does indeed allow you to take advantage of this, in order that people can watch video without "tearing" artifacts due to display memory updates during the monitor scan period, but as I say it does not let you specifically control how many scans each frame gets.

 

It's certainly possible to do this in software, but not, I suspect, off the shelf.

 

If you are more comfortable with a soldering iron than a C compiler, you could come up with a fairly trivial piece of electronics that'd monitor the vertical sync line on the VGA connector and use that to count scans. This might be as simple as using one of those 4000 series fully-decoded decade counters and a rotary switch to control count limit. You might be able to either clamp the RGB lines to black (though be careful about introducing inductance or capacitance that would affect image quality, although suitable high-bandwidth video amplifiers aren't rocket science) or use the information in some other way to control timing. There would be some work involved in achieving correct phase between the electronic event and whatever event you wanted to control with it. This approach has a certain indefatigability about it which may be hard to achieve in software.

 

I think you need to shoot some slides of this monitor and project them, in order to get your results somewhere in the ballpark.

 

I presume you're talking about shooting straight to print stock with this thing? That's balanced for tungsten, isn't it?

 

P

[Keith Walters]

One fly in the ointment:

As far as I am aware, all print stock has slightly different sized perforations from camera negative. I only know this because I once worked on a super high speed 35mm camera, (originally designed for filming nuclear explosions!) and they had to get the sprockets modified to take standard film. The original setup used specially made high-speed negative stock with perfs that matched projector sprockets, so that the original camera negative could be reversal-processed and then used in an ordinary unmodified projector. Hence you would possibly need to get your camera modified at considerable expense.

 

I don't even think you need to do it sequentially. You won't gain yourself much. It gets complicated at this level because of another factor affecting CRT performance that I thought off at about 3am: power supply regulation. As we've seen, both geometry and brightness are affected by the load on the high voltage power supply, which increases nearly linearly with the amount of total light you're asking the tube to put out. It is technically difficult to make an EHT power supply in the tens-of-kilovolts range that is very load stable.

For horizontal scanning and EHT, virtually all VGA monitors use circuitry that is essentially identical to that used in domestic CRT TV sets. An extra winding on the horizontal deflection transformer is used to generate anything between 20-35kV EHT for the picture tube. Rarely is any direct EHT regulation employed, it simply depends on the regulation quality of the main 100VDC rail.

 

This is perfectly adequate for even high-end computer monitors, since the only effect a slightly varying EHT voltage has is to change the picture size slightly, which is why most monitors are slightly underscanned.

 

In a sequential exposure system using a monochrome CRT and switched filters (such as the old Oxberry systems) this is nowhere near good enough. Even minutely different Red, Green and Blue image sizes would cause unacceptable registration errors, so they use separate, tightly regulated EHT supplies. They also need heavy magnetic shielding to avoid the effects of stray AC magnetic fields. Errors that are unoticeable on a 14" monitor could be disasterous on a fifty foot screen (which applies to lots of things of course).

 

So from that viewpoint, a colour CRT would be better since then there would be no image size change caused by switching to the red, green and blue images.

 

However, while I suspect that you are right in suggesting that a full-HD LCD or plasma panel might be a better bet, the problem is that if this scheme really worked, I find it hard to believe that nobody else is already offering this service.

 

1. Modern LCD panels don't strobe and have very low lag times, so you could simply play your video and shoot it at whatever frame rate you wanted without needing any sync setup

2. Most panels are extremely bright, so there should be no trouble using 3 ASA equivalent print stock.

3. The image is all on one plane so you should be able to shoot with the iris wide open once you have the focus right.

 

We've all seen flat-panel displays in movies; the pictures usually look pretty decent, and you'd think with a bit more fiddling they could be made to look even better.

 

So I've got a feeling the main hurdle is going to be modifying the film camera.

[Paul Bruening]

Thanks, Phil and Keith,

 

It appears that a CRT is too much trouble, certainly beyond my technical capabilities. Probably, the reason no one is using that technology for scans is from what you have described. Do you think that those problems are better solved using the IBM DG5 that I have in inventory?

[Paul Bruening]

I keep thinking.

 

One fly in the ointment:

As far as I am aware, all print stock has slightly different sized perforations from camera negative. I only know this because I once worked on a super high speed 35mm camera, (originally designed for filming nuclear explosions!) and they had to get the sprockets modified to take standard film. The original setup used specially made high-speed negative stock with perfs that matched projector sprockets, so that the original camera negative could be reversal-processed and then used in an ordinary unmodified projector. Hence you would possibly need to get your camera modified at considerable expense.

 

As I understand it the Mitchell NC was made to transport up to two layers of film: One print layer and one negative layer to achieve special effects. It has that great big continuous movement cog in the back and a rather large gate, pull-down and pin registration assembly to, in part, do this. I've already run it with one layer of common, processed print stock against my XL2. The registration and steadiness are, to me, profound. My scan NC and record NC are the same design except for the pellicle reflex in the record head.

 

For horizontal scanning and EHT, virtually all VGA monitors use circuitry that is essentially identical to that used in domestic CRT TV sets. An extra winding on the horizontal deflection transformer is used to generate anything between 20-35kV EHT for the picture tube. Rarely is any direct EHT regulation employed, it simply depends on the regulation quality of the main 100VDC rail.

 

This is perfectly adequate for even high-end computer monitors, since the only effect a slightly varying EHT voltage has is to change the picture size slightly, which is why most monitors are slightly underscanned.

 

What is involved in this level of power regulation?

 

Can I simply use underscanning to do this? If I went with a DSLR's shutter bits and a longer exposure to put the exposure control at that place, would a full color scan approach bypass the shorter but difficult digital scan count issue? Would the slow DSLR shutter, underscannning and moderate power control combined with sequential RGB scans resolve all remaining convergence issues? Remember that at the time of their manufacture, these 21" true falts were damn expensive and were the best thing since sliced bread.

 

In a sequential exposure system using a monochrome CRT and switched filters (such as the old Oxberry systems) this is nowhere near good enough. Even minutely different Red, Green and Blue image sizes would cause unacceptable registration errors, so they use separate, tightly regulated EHT supplies. They also need heavy magnetic shielding to avoid the effects of stray AC magnetic fields. Errors that are unoticeable on a 14" monitor could be disasterous on a fifty foot screen (which applies to lots of things of course).

 

So from that viewpoint, a colour CRT would be better since then there would be no image size change caused by switching to the red, green and blue images.

 

I've got Laregraphics LFR Mark V 8K CRT RGB filter wheel scanner. It is as slow as Christmas. But, it was their last and best manifestation of CRT scanning before they dropped the design. It's as heavy as a ship anchor. That must be from the shielding.

 

However, while I suspect that you are right in suggesting that a full-HD LCD or plasma panel might be a better bet, the problem is that if this scheme really worked, I find it hard to believe that nobody else is already offering this service.

 

1. Modern LCD panels don't strobe and have very low lag times, so you could simply play your video and shoot it at whatever frame rate you wanted without needing any sync setup

2. Most panels are extremely bright, so there should be no trouble using 3 ASA equivalent print stock.

3. The image is all on one plane so you should be able to shoot with the iris wide open once you have the focus right.

 

Kinetta used the IBM DG3 and DG5 in his scanner before he gave up the designer's ghost. Lasergraphics also uses some kind of flat panel unit in their scanners. Has anyone seen results from these designs?

 

If you're right about this and I could run my NC against the DG5 with print stock in real time with no sync hassles I could grind out 8 to 12 prints per day. You've got me intersted in the extreme on this possibility. Even if I had to run it at slower frame rates but got solid results I'd be delighted.

 

We've all seen flat-panel displays in movies; the pictures usually look pretty decent, and you'd think with a bit more fiddling they could be made to look even better.

 

I'd love to know this as well.

[Phil Rhodes]

What is involved in this level of power regulation?

 

I suspect the simple answer is "nothing you want to get involved in". There are simpler ways of solving the imaging issues caused by CRT power supplies than rolling your own.

 

If I went with a DSLR's shutter bits and a longer exposure to put the exposure control at that place, would a full color scan approach bypass the shorter but difficult digital scan count issue?

 

You'd still need to control how many vertical retraces you were imaging, unless the number was so high that one more or less didn't make much difference. Even if you take one 8-bit code value as an acceptable deviation (which you probably can't) then you'd need to image 256 scans, which is four and a half seconds a frame on a 60Hz monitor. You probably want to go faster than this, notwithstanding a desire to mitigate EHT regulation issues by using moderate brightness.

 

Would the slow DSLR shutter, underscannning and moderate power control combined with sequential RGB scans resolve all remaining convergence issues?

 

CRT convergence is absolutely never perfect, but it might be good enough for your purposes. The only reason I mentioned it was to explain why people tend to use monochrome CRTs for filmouts (with very careful EHT regulation). You certainly shouldn't do sequential RGB passes in any case, since you will then open yourself up to EHT regulation problems causing convergence issues.

 

So from that viewpoint, a colour CRT would be better since then there would be no image size change caused by switching to the red, green and blue images.

 

Precisely, although a TFT doesn't suffer this in any event.

 

1. Modern LCD panels don't strobe and have very low lag times, so you could simply play your video and shoot it at whatever frame rate you wanted without needing any sync setup

2. Most panels are extremely bright, so there should be no trouble using 3 ASA equivalent print stock.

3. The image is all on one plane so you should be able to shoot with the iris wide open once you have the focus right.

 

They don't strobe much, though some do shimmer a little. I think you'd still want to ensure you were imaging a fixed number of refreshes of a single video frame to a single film frame(with respect to whatever that particular TFT driver does when it refreshes the panel, which will be a characteristic of that particular device). I'd fear not doing this would result in strange double-exposed frames, partial scan imaging, and other undesirable things. You would probably hit the issue of running (presumably) 24-frame video on a display that is inherently incapable of running at an integer multiple of that rate, and even if you sorted that (by finding a rare 72Hz TFT or running everything at 30fps) you will find that you don't, for instance, always get 2 exactly scans of every 30p frame on a 60p display. You could write software to do that, but I'm not aware of anything that does it off the shelf (with the possible exception of a Blackmagic board or something like that which is designed to behave in exactly that way as an essential characteristic). Edit: I believe one of their inexpensive Intensity HDMI cards will do that, and it will display uncompressed movies of the type you might be considering to an HDMI or component monitor using a bundled application. If you can find a way to sync the taking camera to it (edit edit: that you can probably do using one of the analog outputs, which means you need an Intensity Pro), that might be a solution.

 

Wouldn't you also want to run your lens at whatever point gave best optical performance, and fiddle all else to suit?

 

P

[John Sprung]

One fly in the ointment:

As far as I am aware, all print stock has slightly different sized perforations from camera negative.

 

The perforations themselves are the same. The difference is in the distance between them, the pitch. IIRC, it's 0.1866" for camera stocks and 0.1870" for projection stocks.

 

This is one of those strange little differences for which, in fact, a good reason exists. Contact printers use a big sprocket to hold the two pieces of film together and in registration, and the light path to expose the print runs between the flanges of the sprocket and sort of works from the inside out. That's why the print stock has a larger pitch, it's on the outside. It might seem easier to do it the other way, with the print on the inside, but I guess the idea was that you could leave the neg threaded up, and only change the print stock if it's on the outside. Of course you could run the negative both directions, no need to rewind it.

 

 

 

 

-- J.S.

[Keith Walters]

The perforations themselves are the same. The difference is in the distance between them, the pitch. IIRC, it's 0.1866" for camera stocks and 0.1870" for projection stocks.

 

-- J.S.

???

Wouldn't that make the perforations slowly move out of sync with the movie frames?

The distance between the perfs (as in the unbroken stretch of plastic) would indeed be slightly bigger if the holes were slightly smaller, but the centres of the holes must be the same distance apart, surely.

[Hal Smith]

Convergence is an issue of three electron gun CRT's and CRT projection sets. Since the guns are in a triangular arrangement in direct viewing tubes and a linear arrangement in projection there are parallax and/or tilt (trapezoidal) geometry issues that are solved with dynamic convergence circuitry.

 

All modern CRT's have a single electron tube with phosphors arranged on the inside face of the tube in vertical strips therefore no parallax, no tilt issues. Best known of course is the Sony Trinitron tube.

 

If it were up to me I'll build a DIY kinescope setup around a high-end plasma like my 58" Samsung. The better plasmas have a large set of gamma and RGB adjustments that would enable tuning the display for best transfer function from video to film.

[Paul Bruening]

Hey Hal,

 

I've already got this 3840 IBM DG5. Have you seen one in operation? Pretty stunning. I'm crossing my fingers that I can make it work since the CRT option is looking complicated well beyond my technical skills.

 

http://en.wikipedia.org/wiki/IBM_T221

 

You know, the T221 (DG5) runs off two DVI outputs at 48 Hz. What's the chances of timing that natural refresh rate with a camera's native cycles? Could an overcooled quad core mobo run it off mirrored 3.0 SATAs?

[John Sprung]

Wouldn't that make the perforations slowly move out of sync with the movie frames?

 

No, the frames also get printed farther apart by that same tiny amount, 0.0004". There are no frames on the print stock until it gets exposed. The difference has to do with the developed neg and raw print stock being bi-packed around a sprocket, sort of like how one wrap around a roll of film will be just slightly longer than the one before it.

 

Perhaps Dominic and Robert can add something, this is their turf.....

 

 

 

 

 

-- J.S.

[Dirk DeJonghe]

You can order print stock with 'positive' = KS or 'negative'= BH perforations from Kodak and Agfa and I suppose Fuji too. I can record to positive stock on my Lasergraphics P3 CRT recorder. This is why I ordered short pitch BH standard perforations for the print stock. It runs through the projector just fine, with just a little bit more noise. It takes about one hour to record one minute of film.

 

I am sure you will get a picture if you put enough light on the film, but to get a really good picture that matches or exceeds anything else on the market, you will have a long road ahead, CRT machines have evolved over many years, I started with a Solitaire Cine3FLX 15 years ago and the difference with a modern CRT such as the P3 is dramatic.

 

Cinevator uses a Arri-like film movement and a DLP-type image source. The machine also costs 1million Euros plus annual maintenance fees.

[Phil Rhodes]

Could an overcooled quad core mobo run it off mirrored 3.0 SATAs?

 

Yes, but that's not the problem, the problem is getting everything to happen at the right micro-instant. It can be done, but it isn't about raw power.

 

P

[Keith Walters]

Here’s an even crazier idea.

 

Most consumer LCD panels only have 6-bit digital to analog conversion for the red, green and blue pixel drivers. Although in general this works a lot better that you might expect for normal TV viewing, it may be problematic if you are trying to transfer HD video to film

Most panels (and all cheaper ones) also use a cold-cathode fluorescent lamp to provide the LCD backlighting, which simply runs at maximum brightness while ever the display is running.

 

With the recent demand for greater energy efficiency, manufacturers have tried to develop ways to turn down the backlighting. A major problem with “cold cathode” tubes is that their cathodes are not really cold at all; they depend on the discharge current to keep them glowing. The upshot of this is that you can only dim them a relatively small amount before they become unreliable, and running them at reduced brightness also shortens their life.

 

More recent up-market models have started to use LED backlighting. LEDs are particularly good for this application because their brightness can be controlled over a very wide range, and virtually instantly, with no colour shift. For the casual experimenter, they would also be a lot safer to play around with than a fluorescent backlight. Plus they can be switched on and off millions of times without deterioration (unlike fluorescent tubes)

 

So the problem as I see it is that you want to transfer a 10-bit (say) digital video signal through a system that only has 6-bit resolution. That is your original 1024 possible red, green and blue levels are reduced to 32 possible levels.

 

Suppose we break the 10 bit number representing each red, green and blue pixel into two parts: bits 0 to 5, and bits 6 to 9. Both parts of the number are then numerically massaged so that they both occupy the range 0 to 255, which is the sort of levels expected by most consumer panels In that case the 10-bit number 111111 XXXX would come out as 1111 1111 ie 255. (The XXXX part means those bits have no effect on the outcome) as would the 10-bit number XXXXXX 1111.

What we do then is sequentially display the video frames, first using the lower 6 bits, and then the higher 4 bits.

 

But here’s the sneaky bit. When the pixels derived from bits 0-5 are being displayed, we turn down the LED backlight to one/sixteenth of the level we will be using when the upper 4 bits are being displayed. The result will be a peak white of one sixteenth the normal level broken up into 64 levels, and then the full peak white also broken into 64 levels.

 

In theory this should give a 12-bit equivalent greyscale.

 

It would probably be easier and more accurate to simply vary the LED brightness by changing the duty cycle, and in fact you could use this to control the exposure. For the upper 4 bits you might expose for 1 second, and for the lower six, about 60 milliseconds.

That way the sequence would be: LEDs off, feed in lower six bit picture to th eLCD panel, pulse LEDs on for 60mS, LEDs off, feed in upper four bit picture, pulse LEDs for one second, LEDs off, advance film, feed in lower six bit picture etc etc

 

Would this work? Who knows.

 

However, if you're careful, you could always put the panel back the way it was and watch TV on it!

 

One thing in your favour: modern flat-panel displays are an order of magnitude easier to work on than the ones of just a few years back, when you needed a special and absurdly expensive jig to manipulate those horribly expensive and fragile panels.

 

I was just looking at a 32" Sany LCD TV today. It only weighs 11.5kG, less with the foot removed. You literally could hang it on a wall like a picture. A few years back you needed a special mounting bracket that could only safely be mounted on a brick wall, and some of them cost more than a quite decent CRT TV!

[Phil Rhodes]

32 possible levels.

 

64. But you can find 8-bit panels relatively easily. You'd need to be particularly careful with 6- or 7-bit panels that might dither.

 

This is of course one reason CRTs might be less of a problem. There have even been 10-bit graphics cards with which to drive them.

 

P

[Keith Walters]

64. But you can find 8-bit panels relatively easily. You'd need to be particularly careful with 6- or 7-bit panels that might dither.

 

This is of course one reason CRTs might be less of a problem. There have even been 10-bit graphics cards with which to drive them.

 

P

Right, 64.

As for 8-bit panels, that's still not enough bits though, so you'd still probably have to use some variation of this technique. I don't see why dither would be a problem.

 

The main advantage of my scheme is that it would allow you to use relatively cheap consumer 1920 x 1080 panels. If you could get enough film cameras, and could make the system work at 1 fps, a "suite" of 7-8 of these could turn out a full-length feature film in about 8 hours as a set of 20 minute reels.

[Paul Bruening]

Doing the best I can to take all these differing technical viewpoints in my best guess is to go with a low tech solution:

 

Put a 24V servo-capping shutter in front of the Mitchell's 75mm Cooke lens. Run print stock through it since the slow speed will accommodate the shutter's slight variations in open time. Run it in front of the DG5 since there are little to no issues when exposed for 2-5 seconds. Run the stepper-single advance motor and shutter off a CNC controller in time with AE or PS display of the 3200 x 2400 digitally squeezed images for eventual scope output. Live with the fact that it will take a month to grind out 3 1st gen positive prints or make one negative on the recorder and contact print the rest.

 

Does slow and simple eliminate all of those timing issues?

[Keith Walters]

Doing the best I can to take all these differing technical viewpoints in my best guess is to go with a low tech solution:

 

Put a 24V servo-capping shutter in front of the Mitchell's 75mm Cooke lens. Run print stock through it since the slow speed will accommodate the shutter's slight variations in open time. Run it in front of the DG5 since there are little to no issues when exposed for 2-5 seconds. Run the stepper-single advance motor and shutter off a CNC controller in time with AE or PS display of the 3200 x 2400 digitally squeezed images for eventual scope output. Live with the fact that it will take a month to grind out 3 1st gen positive prints or make one negative on the recorder and contact print the rest.

 

Does slow and simple eliminate all of those timing issues?

How are you going to add the sound?